Learning the Approach During the Short-loading Cycle Using Reinforcement Learning